No abstract

In recent years, the massive growing of the usage of Internet of Things (IoT) applications expose different challenges in centralized cloud computing paradigm such as; network failure and inadmissible latency for real-time services. The fog layer is added to address these challenges. The nodes in fog computing layer are similar to cloud node, except that it is nearby the IoT devices. Fog node is added to supply IoT devices with the desired resources with low delay. Hence, the performance of IoT application is based on the task scheduling strategy in cloud, or fog computing systems. Hence, the scheduling strategy should maximize resource utilization and increase the resources availability. Most of the previous scheduling methods are suffer from centralization, which consequently represents a performance bottleneck and a single point of failure. Also, some of these scheduling methods ignore the urgency type of the services, which consequently will not be suitable for real-time services. This paper proposes a new load balancing model, which has two main features. The first feature is the dynamic allocation quota for the resource allocation. The second feature is decentralization in fog resource management. In another word, DLBRT manages the IoT service requests based on the service urgency level with maintaining the load among the fog node balanced. The dynamic load balancing strategy in DLBRT is based on redistribution of the less-urgent service requests to the lowest fog node load. In another word, the redistribution of the service requests is depending on the urgency of the real-time service and the workload in every fog. Finally, a simulation model is created to evaluate DLBRT in a fog-cloud colony. Also, the proposed scheduling model is examined with four scheduling models, namely; The FCFS, Max-Min, Real-Time Efficient Scheduling (RETS), and Based Autonomic Task Scheduling (PBATS). We demonstrate through thorough simulations that the performance metrics of turnaround time, waiting time, throughput, and task failure test are enhanced by our suggested approach.